On 4/9/07, Ken Restivo <ken@xxxxxxxxxxx> wrote:
Well, I have a great deal of hope for truly SSEing many of the plugins. In particular, if it were possible to stop using the "Flush_to_zero" macro and just turn on SSE's builtin support for flushing denormals to zero for plugins that care might be a big win - but that may have audible effects elsewhere.
As for comb_run - which accounts for a large precentage of tap_reverb's runtime, for example - SSEing it cuts the number of ops by a lot, but as it has poor locality of reference so it may not actually SSE well. My current test case is the parametric EQ which spends a lot of time there, but also has some gnarly setup....
only answer is to try it and benchmark... may have some results by the end of the month.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Mon, Apr 09, 2007 at 08:18:10AM -0700, Mike Taht wrote:
> At present the major bottlenecks left for ardour and jack are very much down
> in the noise floor. A typical user now spends more cpu time in plugins than
> in those two core programs. (thus I've been oprofiling a few plugins
> heavily, among other things, now have a SSE'd comb_run routine... and
> hopefully will announce some sped up plugins soon)
>
That would be a HUGE help for me.
Hint hint: the CAPS and TAP plugins would be especially nice to have optimized.
(Yes, I tried compiling them with SSE enabled, using GCC 4.1, and experienced *higher* CPU usage that way.)
Well, I have a great deal of hope for truly SSEing many of the plugins. In particular, if it were possible to stop using the "Flush_to_zero" macro and just turn on SSE's builtin support for flushing denormals to zero for plugins that care might be a big win - but that may have audible effects elsewhere.
As for comb_run - which accounts for a large precentage of tap_reverb's runtime, for example - SSEing it cuts the number of ops by a lot, but as it has poor locality of reference so it may not actually SSE well. My current test case is the parametric EQ which spends a lot of time there, but also has some gnarly setup....
only answer is to try it and benchmark... may have some results by the end of the month.
- -ken
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
iD8DBQFGGmnTe8HF+6xeOIcRAvdTAKCpUK8TR6rmCMJu84IXl+aAV6oqvACeLSTo
IM8wC/gFZCvvTs40gsLBAJg=
=IFbH
-----END PGP SIGNATURE-----
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@xxxxxxxxxxxxxxxxxxxx
http://lists.linuxaudio.org/mailman/listinfo.cgi/linux-audio-user
--
Mike Taht
PostCards From the Bleeding Edge
http://the-edge.blogspot.com
_______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx http://lists.linuxaudio.org/mailman/listinfo.cgi/linux-audio-user
